BACKGROUND: The effectiveness of computer-based cognitive training (CCT) remains controversial, especially in older adults with neurodegenerative diseases. AIMS: To evaluate the efficacy of CCT in patients with Parkinson's disease and mild cognitive impairment (PD-MCI). METHODS: In this randomized controlled trial, 53 patients were randomized to receive CCT delivered by means of CoRe software, traditional paper-and-pencil cognitive training (PCT), or an unstructured activity intervention (CG). In each group, the intervention lasted 3 consecutive weeks (4 individual face-to-face sessions/week). Neuropsychological assessment was administered at baseline (T0) and post-intervention (T1). Outcome measures at T0 and T1 were compared within and between groups. The Montreal Overall Cognitive Assessment (MoCA) was taken as the primary outcome measure. RESULTS: Unlike the PCT group and the CG, the patients receiving CCT showed significant medium/large effect size improvements in MoCA performance, global cognition, executive functions, and attention/processing speed. No baseline individual/demographic variables were associated with greater gains from the intervention, although a negative correlation with baseline MoCA performance was found. CONCLUSION: CCT proved effective in PD-MCI patients when compared with traditional PCT. Further follow-up assessments are being conducted to verify the retention of the gains and the potential ability of the tool to delay conversion to PD-dementia. Trial registration number (ClinicalTrials.gov): NCT04111640 (30th September 2019).